Biopizysicai Chemistry 41 (1991) 263-276 Elsevier Science Publishers B.V., Amsterdam
263
BIOCHE 01596
Interfacial thermodynamics of protein adsorption and ion co-adsorption. III. Electrochemistry of bovine serum albumin adsorption on silver iodide J.G.E.M. Fraaije *, W. Norde and J. Lyklema Department of Physical and Colloid Chemistry, Agricultural University, Dreijenplein 6, 6703 HB Wageningen (Netherlands)
(Received 7 January 1991;accepted in revised form 19 April 1991)
Abstract An experimental analysis of charge regulation in protein adsorption is presented. The model system consists of colloidal particles of the slightly water soluble salt silver iodide as the adsorbent and the protein bovine serum albumin as the adsorbate. Protein adsorption experiments corroborate earlier findings that albumin adsorbs maximally close to the isoelectric point of the protein. The adsorption is reversible with respect to proteinprotein exchange. The charge regulation is studied by novel potentiometric titrations. The Galvani potential of the adsorbent, partially covered with protein, is varied by the addition of AgNO,/Kl while the pH is kept constant by means of a pH-stat. It is shown that the ion co-adsorption is a linear decreasing function of the blank surface charge density. The results are consistent with thermodynamics: for the first time a few phenomenological linkage relations between the ion co-adsorptions and chemical potentials are verified experimentally. The charge regulation is interpreted in terms of a contact layer model, which explains the ion co-adsorption by compounded ion exchange equilibria in the small layer of atomic contact between adsorbed protein and surface. Keywords: Proteins; Thermodynamics;
Adsorption;
Ion co-adsorption;
1. Introduction
In a series of papers we addressed the interfacial thermodynamics of protein adsorption, ion co-adsorption and related issues. We recently have discussed solubilization of protein in reverse micelles [l], protein ion exchange chromatography [2,3] and the thermodynamics of ion binding
* To whom correspondence should be addressed at AKZO Corporate Research Laboratories Arnhem, Department Applied Mathematics, P.O. Box 9300, 6800 SB Arnhem, Netherlands.
0301-4622/91/$03.50
Ion exchange; Silver iodide; Bovine serum albumin
to proteins in solution [4]. In this report we apply interfacial thermodynamics to a model experimental system, namely adsorption of bovine serum albumin (BSA) on the colloid silver iodide (AgO. The model experimental system is chosen such that the Galvani potential of the surface can be controlled, independently of the pH of the solution. Albumin (human and bovine) adsorption in relation to ion co-adsorption has been studied before, using various techniques: adsorption on porystyrene (ion co-adsorption measurement by electrophoresis [5,6] and radiometry [7]) and hydrophobic copolymers (ion co-adsorption mea-
0 1991 - Elsevier Science Publishers B.V. All rights reserved
264
J.G.E.M. Fraaije et al. / Thermodynamicsof ion co-adsorption,Ill
surement by titration [S]). The experiments show that in many instances ion co-adsorption is not solely determined by electrostatic factors. It is well known that protein adsorption may be accompanied by (partial) unfolding and that the unfolding may lead to a shift in ion binding properties (see review [9J and references therein). Especially albumin is an interesting protein in this respect because it is a labile molecule, which has been shown to unfold on various negative surfaces in roughly the same way [9]. For pH close to the isoelectric point (i.e.p., about pH 4.7), where the protein is Ieast labile in solution, the molecure unfolds minimally, leading to relatively high adsorbed amounts of 2-3 mg/m2. For pH > i.e.p. or pH < i.e.p. the protein is more labile and unfolds to a larger extent, and the more so, the further the pH is away from the i.e.p., leading to relatively smaller adsorbed amounts. The coIloid AgI is one of the classical model systems from colloid chemistry (for general properties see review [lo]>. The colloid can be prepared as a sol or a suspension of precipitates (particle size O.l-10pm). The precipitate lends itself to titration, by adding AgNO, and/or IU solution, without unwanted irreversible additional flocculation. Such titrations are usually analyzed in terms of the surface charge density rO, defined as (eq. [15] from [lo]):
where F is the Faraday constant and ri is the adsorbed amount of i in mol per square meter, a, is in Coulomb per square meter. Note that in the present paper, as in the previous papers [l-4] and [lo], we prefer discussion of interfacial quantities in terms of adsorption of electroneutral components. However, for purpose of interpretation one may identify a, with the excess adsorption of Ag+ ion minus that of I- ion. It is not possible to discriminate between a positive adsorption of one of the components and a negative adsorption of the other. The second experimentally observable, by the use of a AgI electrode, is the silver ion activity, expressed as pAg = -log(a,,+), which can be
related to the interfacial Galvani potential A@ through Nernst’s Law: A@= @a+ T
ln( uAg+)
(4
where a,, is a constant. By convention, the value of G0 is chosen such that A@ is zero in the zero point of charge pAg, of the blank in indifferent electrolyte (KNO,). Silver iodide has been titrated in the presences of various adsorbates, e.g. alcohols [ll], polyvinylalcohols {12],tetralkylammonium nitrate salts [13], oligo- and polypeptides [14]. It is generally found that the adsorbates shift the point of zero charge to lower pAg levels, and that they decrease the differential electric capacitance. The former feature is explained by displacement of oriented surface bound water molecules by hydrophobic moieties of the adsorbates, the latter by reduction of the ratio Es/d, between the dielectric constant E, and thickness d, of the inner double layer part. Here, we are interested in the titration of AgI in the presence of BSA, which may cause an additional effect. Namely, BSA contains weak acid and base side groups so that adsorbed BSA may respond to an imposed shift in Galvani potential by a shift in the acid/base equilibria, i.e. the ion co-adsorptions of acid/base and AgNO,/IU may be dependent on the charge and/or potential of the surface. In our model experiment we monitor the extra uptake or release of acid/base with a pH-stat. The pH-stat introduces two additional observable quantities, namely the pH and the acid/base surface charge density a,,, defined as the excess adsorption of acid over that of base: (3)
As before, the adsorbed amounts are those of the electroneutral components. Negative adsorption of acid is experimentally undistinguishably from positive adsorption of base and vice versa. Bovine serum albumine does not contain hydroxyl binding sites, nor does AgI adsorb acid or base (acid or base binding is negligible for pH G 7, see below). We therefore interpret gab in terms of extra
J.G. E.M. Fraaije et al. / Thermdynamics
of ion co-adsorption, III
265
proton charge (extra with respect to the protein in solution) of the acid/base groups of the adsorbed protein. The AgI titration with pHstat allows us to measure in a single experiment the surface charge of the AgI colloid v0 and the acid/base surface charge density a,,, both as a function of the pAg (or A@) and pH. We analyze the results in two complementary ways. First, the findings are discussed with the thermodynamic method [2]. We will show that various phenomenological linkage relations between the ion co-adsorptions can actually be verified experimentally. Second, the observed behavior is explained with the contact layer model we introduced before [3]. The contact layer model explains ion co-adsorption by compounded ion exchange equilibria in the small layer of ‘atomic’ contact between adsorbed protein and surface. The model allows interpretation of various interesting relations, e.g. that between gO,a,, and the area of the contact layer.
mates of the specific surface has been found before [lo-141; no satisfactory explanation has been given so far. In this report we prefer using the area from the dye adsorption measurement. In the Results section (Section 3) we will discuss this choice in some more detail.
2. Experimental
Typically, O-5 ml of protein solution (containing O-3 mg of BSA in 0.1 M KNO,) were added to 5 ml of we11dispersed AgI precipitate (containing 1.7 g Agl in 0.1 M KNO,) in 10 ml Sybron/ Nalge polycarbonate centrifuge tubes. After the total volumes of the tubes had been adjusted to 10 ml with 0.1 M KNO, solution, the tubes were stoppered with polyethylene caps and vigorously shaken for 10 seconds on a whirl mixer. The tubes were then rotated end over end for 16 hours to reach steady state (more than 90% of the final protein adsorption was already reached in less than a quarter of an hour). Next, the tubes were centrifuged for 20 minutes at 20,000 rpm in a Beckman JA-21 centrifuge equipped with a JA-21 rotor. The protein concentration in the supernatant was determined by measuring the absorbance at 280 nm.
2.1 Preparation of siluer iodide precipitate We closely followed previously published procedures [lo-141. Two batches of AgI precipitate (A and B) were prepared as follows. Two liters of 0.105 M AgNO, were slowly added to 2 1 of 0.1 M KI at a rate of about one drop per 10 seconds. The reaction vessel was shielded from light and continuously flushed with carbondioxide-free nitrogen in order to prevent the formation of silver oxides. After extensive washing with de-ionized water the precipitates were stored at pAg z 6 in the dark under nitrogen atmosphere. Measurement of plateau adsorption of methylene blue yielded for the specific surface areas 0.25 m*/g for precipitate A and 0.30 m2/g or precipitate B. The latter precipitate was used for the titrations and protein plateau adsorption measurements. The point of zero charge of precipitate B was 5.80, very close to the literature value 5.67. From capacitance measurement in the point of zero charge in dilute KNO, (1.1 mM) we calculated a specific surface area of 0.88 m*/g (for method see [lo]). The discrepancy between the two esti-
2.2 Preparation of BSA Bovine serum albumin was purchased from Sigma (type VI, y-globuline and fatty acid free). The number of sulfhydryl groups per protein molecule was initially between 0.3 and 0.4. We increased the sulfhydryl content by a purification method [15]; after purification the number was between 0.85 and 0.88. The sulfhydryl group was subsequently blocked by carboxyamidation, either with normal or radioactive iodoacetamide (Cl41 [ 151.The protein, freed from salt and freeze-dried, was stored at -20°C. 2.3 Adsorption isotherms
2.4 Protein-protein exchange First, 5 ml of unlabelled BSA (2 mg/ml) was added to 25 ml of dispersed AgI precipitate (sorid contents 0.37 gram/ml) in a 50 ml Schott/Duran flask. The mixture was then vigorously stirred for 16 hours using a magnetic stirrer. After the pro-
266
J.G.E.M. Fraaije et al. / Thermodynamics
tein adsorption had been determined, a small amount of radioactive BSA was added, less than 5% of the total amount of protein already present, and the stirring was continued. At selected time intervals 0.5 ml aIiquots were taken from the dispersion and centrifuged in a top desk centrifuge in 1 ml polyethylene vials. The concentration of radioactive BSA was determined with a scintillation counter. The adsorptions of radioactive and unlabelled BSA were calculated from the respective depletions.
of ionco-adsorption,
III
and KI solution in 0.1 M KNO, solution. The effect of pH on the E, values was negligible ( < 1 mV>within the pH range we studied (4 < pH G 71, conversely, the pH drift during the Agl electrode calibration was also negligible (< 0.01 pH unit in unbuffered pH 5.5 solution). The Agl solubility product pS was 16.03-16.06, slightly lower than the literature value 16.07 [lo]. The pH electrode was calibrated with commercial buffers (Titrisol) of pH 4 and pH 7. The pH and AgI electrode calibrations were repeated directly before and after every AgI titration.
2.5 Titration, instrumentation 2.7 Titration, procedure The titrations were carried out in a thermostatted (20 ’ C) Schott titration vessel (volume 1.50ml). The vessel was equipped with four AgI electrodes fitted in one holder, a Schott pH glass electrode and a so-called Van Laar salt bridge (see reference 120 in [lo] for original details) with resistance 700 k0. The reference chamber contained saturated KC1 and a Schott Ag/AgCl reference electrode. The electrode potentials were measured with a multichannel voltage meter (HP 3497A Data Acquisition/Control Unit) after they had been converted to low impedance signals with a home-built impedance transformer. The titrants (AgNO, and KI both 0.01 Iw in 0.1 M KNO,, HNO, and KOH both 0.025 M in 0.1 M KNO,) were added with a Metrohm 655 Dosimat. Homogenization of the concentrated dispersion (about 30% by weight) was achieved by stirring vigorously at the bottom and top of the vessel. The experiments were automated by means of a HP 85 microcomputer. Every five minutes, the computer program collected the data from the electrodes, averaged the signals from the four independent AgI electrodes, decided whether equilibrium had set in (tolerated drifts 0.001 pH/min and 0.001 pAg/min, tolerated spread between the AgI electrodes 0.02 pAg), ordered the burets to add small aliquots of titrant (O.Ol0.1 ml) if necessary, and stored the data. 2.6 Titration, calibration The AgI electrodes were calibrated by short titrations (from 10e5 to 5 . 10e3 M) with AgNO,
First, wet AgI precipitate was added to the titration vessel to a final amount of 28-30 gram, dispersed in 100 ml 0.1 M KNO, and adjusted to pAg 5.5 and selected pH. Then a blank titration was carried out, up to pAg 10.5, in nine steps by successive addition of KI titrant. Next, the titration was reversed in one step by the addition of AgNO, titrant. The initial pAg was always recovered within 0.02 pAg unit. Then 2 mg BSA was added as 1 ml freshly prepared protein solution in 0.1 M KNO, and also at the selected pH level (* 0.001 pH unit). Any pH changes in the titration vessel were first corrected by the pHstat and then the KI titration was repeated. During the titration, the pH-stat continuously monitored the pH, and corrected drifts of pH by addition of the acid and/or the base titrant. In the experiments described, the pH-stat tolerance was set to (in units of pH): 0.001 (pH 4), 0.003 (pH 5) and 0.01 (PH 6). The total procedure for one titration curve took 6-8 hours, the scheme was repeated several times with increasing loads of BSA.
3. Results 3.1 Adsorption isotherms Some typical adsorption isotherms are presented in Fig. 1 as a function of initial pH and pAg. The specific surface area we used for the calculation of the adsorption values in mg/m’
J.G.E.M. Fraaije et al. / Thermodynatnicsof ion co-adsorption,III
was from the dye plateau adsorption measurement (0.25 m*/g for precipitate A), as explained in the experimental section. The protein plateau adsorption values (see also Fig. 2) range from l-2.5 mg/m*, in fair agreement with commonly accepted plateau adsorption values of albumin [9]. If we would have used the surface area (0.88 m*/g) from the capacitance measurement, then the plateau values would have been in the range 0.3-0.8 mg/m’, which is obviously too low. The adsorption isotherms of Fig. 1 reveal that up to about 75% of the plateau adsorption the bulk concentration is below 0.001 mg/ml (detection limit). The isotherms are for four different combinations of initial pH and pAg, including adsorption of (almost) uncharged protein on (almost) uncharged AgI (pH 5, pAg 6.1) and at the other extreme adsorption of negative protein on (very>negative AgI (pH 7, pAg 10.1). The adsorption clearly remains of the high affinity character, even for the case where negative BSA adsorbs on negative AgI. We found that after protein adsorption the equilibrium pH and pAg levels where slightly shifted (less than 1 pH or pAg unit) from the initial values, indicating that the protein adsorption was accompanied by ion co-adsorption. In
267
r, Img /m?
20
!
15-I
10-l
05
L-m-7 PH
Fig. 2. Plateau adsorption of BSA on Agl vs. equilibrium pH and pAg (interpolated, see text) in 0.1 M IWO,.
general, these shifts were similar to those measured more directly in the titration experiments described below. 3.2 F’fateau adsorption
0
I
0.05
,
0.1
I
0.15
I
0.2
I
0.25
cp(mglml) Fig. 1. Adsorption & of BSA on AgI vs. equilibrium concentration cp in 0.1 M IWO,. Initial pAg, pH values: (v ) pAg 6.1, pH 7; (0) pAg 10.1, pH 7; (A) pAg 6.1, pH 5; and (0)
pAg 10.1, pH 5.
In a separate experiment we varied the initial pH and pAg (in a matrix of 11 points variation in pH vs. 4 points variation in pAg, i.e. in total 44 points), while the total amount of protein was held constant (2.5 mg>. The equilibrium protein concentration in bulk varied with adsorbed amount from 0.11 mg/ml to 0.2 mg/ml, this is in the plateau regime (Fig. 1). These adsorption values were (graphically) interpolated with respect to the equilibrium pH and pAg values, to values for constant pH and pAg. The results are shown in Fig. 2. The curves show that the adsorption is a maximum at pH 4, close to, but siighly lower than the i.e.p. We noted already in the Introduction that Albumin adsorbs on a variety of negative surfaces with maxima close to the i.e.p. of the protein (pH 4.7) [91, our results confirm this. There is a small effect of the surface charge:
268
J.G.E.M. Fraaije et al. / Thermodynamics
of ion co-adsorption, IfI
_________________--------k
,OO
*L
per, %
O
@@A-A-
6060-
/
'o?&+&?d&@~ ipO&----o zo-
0
I
I I20
60
1 240
1SO
I 420
I 360
I 300
n y
I 1420
t(mn)
Fig. 3. Protein exchange ratio p.e.r. vs. exchange time. 0.01 M KNO,, pAg 10: (0) pH 6, cp = 0.08 mg/ml; mg/ml; ( a) pH 4, cP = 0.03 mg/ml.
we find that the adsorption is slightly higher when the charges of protein and surface are opposite.
rapid exchange to p.e.r. E 30-40% in the first few minutes, and a slower exchange to p.e.r. E 100% in the following hours. Moreover, there is Little influence of the pH. This demonstration of reversibility is important: it is a prerequisite to verify the thermodynamic linkage relations in the Discussion.
3.2 Protein-protein exchange experiments As a measure of the protein exchange we used the protein exchange ratio (p.e.r.) = (r/C)radioactive BSA ' (C/r)u"labelled BSA' The
(x) pH 5, cp = 0.07
3.3 Titrations
de-
The titration curves are shown in Fig. 4. Adsorption of the protein is reflected in the shifts in u,, and a,, with respect to the blank values. Note
pendence of p.e.r. on the medium conditions and time is depicted in Fig. 3. The results clearly indicate that the adsorption is reversible, with a
-100(a)
1 I
pH4
I
I
I
I
/
1
I
I
I
I
I
I
6
7
6
9
10
11
6
7
6
9
10
11
PAg
PAg
Fig. 4. Titrationof A&BSA complex in 0.1 M KNO, at fixed pH, r0 is surface charge density of AgI (eq. [l]), vaah is acid/base surface charge density (eq. [3]). pH 4 and pH 5: (0) blank; (x) r,,= 0.47 mg/m2; (a) rP = 0.93 mg/m2; and (0) rP = 1.4 mg/m’,
ph 6: (0) blank; (X) r, = 0.33 mg/m’;
(A) r, = 0.67 mg/m*; and (u) r, = 0.93 mg/m*.
J.G.E. M. Fraaije et al. / Thermodynamicsof ion co-adsorption, Ill
u,JmC/m*
269
1
soo - (4
I 6
7
e
9
PA9
I IO
I 6
I 11
I 7
a
9
PA9
IO
J 11
Fig. 4. (continued).
that in these experiments the protein adsorption levels are beIow the plateau adsorption values and that consequently the protein concentration in solution is negligibly low. We may therefore identify the charge regulation with the response of adsorbed protein only and neglect contributions from titration of the protein in solution. The blank curves of a0 for the different pH levels are identical, illustrating the absence of base/acid adsorption on the bare AgI surface. This is confirmed by the (undetectably) low
acid/base adsorption craabin the blank titration, The shapes of the blank r0 curves are in good agreement with published titration curves of Agl precipitates in 0.1 M KNO,, although the absolute values from our measurements are about a factor 3 higher [lo]. The explanation lies in the discrepancy between the specific surface area from capacitance measurement (often used in the older work [lo], 0.88 m2/g for the precipitate B), whereas here we prefer to use the specific surface area from the dye adsorption measurement (0.30
ff.b(mC/m*
)
$100 if) t
PH 6
Fig. 4. (continued).
J.G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption, Iii
270
*r,
40
(a)
J
.1 -0.3
*
4c
40
-0.2
-0.1
0
*WV)
r,,
(4
(C)
x pH5
B
2c
A
f
t-
C
pH5
-at
-0.3
-03
I
-20
0
13
/
-a2
A+(V)
,I
-0.1 *O(V)
Ar.
*r,,
40
4C
(4
2c
(
20
t
01
0
B
*
a
P%
-20
1
-0.3
-92
-01 AOCV)
-“.a
-a2
-01
0
Ablv)
Fig. 5. Ion co-adsorption numbers Ari vs. interfacial Galvani potential A@. pH 4 and pH 5: (0) blank; (*) r = 0.47 mg/m2; (A) = 0.93 mg/m*); and (0) r, = 1.4 mg/m’. pH 6: (0) blank; (*I r, = 0.33 mg/m’; (A) r, = 0.67 mg/mP; and ( q ) T, = 0.93 mg/m*.
r,
J.G.E.M. Fruaije et ul. / Thermodynamicsof ion co-adsorption,III
m2/g for the precipitate B). Note that we do so in order to make the BSA plateau adsorption levels conform to the literature values. We find that in general the more negatively charged protein (pH 6) makes a0 more positive, and that the more positively charged protein (pH 4) makes a, more negative (except for pAg > 8.5). Also, in all cases a more negatively charged surface makes the protein more positive. These effects are clearly a result of electrostatic induction. Note however that for pH 4 (Fig. 4a) there is a common intersection point (c.i.p.) in the a, curves (around pAg 8.51, such that for pAg > c.i.p. the surface, partially covered with positively charged protein, is less negatively charged than the blank. A second feature is that the differential electric capacitance C (= b&IA@ = - 17.19 auJ apAg in F/m’> is not much affected by the protein adsorption. Comparison with effects of other types of adsorbates [lo-141 shows that in this respect BSA behaves exceptionally (see Introduction). In general, the effect of an adsorbate is to reduce C significantly, which is explained by a possible increase in the thickness of the Stern layer and/or lowering of relative dielectric constant of the Stern layer. The implicit assumption is then that the Stern layer itself cannot adjust in charge, i.e. is not titratable. Here, the adsorbed protein readily adjusts its charge through titration of the acid/base groups in the contact zone between protein and surface, which is the region dominating C.
271
(a) pH4
-120
-80
-40
40
0
dcmom?,
b) PHI
-120
-80
-40
0
dhlC/m2) A& + Ar,,
4or
3.4 ion co-adsorption numbers A slightly different presentation of the titration results, especially suited for the thermodynamic and model interpretation, is in terms of the ion co-adsorption numbers of the acid/base, Arab, and AgNO,/KI, Ar,. These are the numbers of ions extra adsorbed or desorbed per adsorbed protein molecule. They are defined through: 1 Ar,=I;T(q,-qb) P
Fig. 6. Total ion co-adsorption Ar, + Ar,, vs. surface charge density of the blank I$. Symbols as in Fig. 5.
1. G.E.M. Fraaije et al. / Thermodynamicsof ion co-adsorption,III
272
where gob is the surface charge density of the blank. The co-adsorption numbers were calculated from the original titration curves Figs. 4 through graphical interpolation. The results are shown in Figs. 5 (note that we have also converted pAg to the Galvani potential by the use of eq. [2] and pAg, = 5.67 for the blank). We recall from the proton titration experiments by Tanford [121 that the isoionic point of BSA in solution is about pH 5.5-5.6 and the isoelectric point is about 4.7. The proton binding number is about 29-34 for pH 4, 6-7 for pH 5 and -3 for pH 6 (lower numbers measured in 0.08 M KCI, higher numbers in 0.15 M KCl>. The maximum positive proton binding number is about 96, i.e. this number of protons may be bound in excess to the isoionic state [4]. This implies that the maximum number the protein can co-adsorb is about 62-67 at pH 4, 89-90 at pH 5 and 99 at pH 6. The co-adsorption curves show that in the present case the co-adsorption levels are far below the maximally attainable levels, indicating that in principle the protein has ample capacity to adjust in charge by the extra uptake or release of protons. However, it is possible that the saturation observed near Arab G 30 for pH 4-5 (Figs. 5b,d) indicates that the effective maximally attainable co-adsorption numbers are smaller than the theoretical values we calculated. This may be explained by the contact layer
Table
PHI 0.47 0.93 1.4 PHI 0.33 0.67 1.0
4. I Thermodynamic interpretation
Generally speaking, thermodynamics predicts that co-adsorption numbers and chemical potentials are related through various Maxwell (or linkage) equations. These equations arc easily dcrived from a set of coupled Gibbs-Duhem relations, as we have shown in [2]. In the Appendix the theory is applied to find the linkage relations between the co-adsorption numbers Ar, and Arab
analysis of Ar, + Ar,, vs. uCv(Fig. 6) R2
Intercept
(ms/m*) pH4 0.47 0.93 1.4
4. Discussion
1
Regression
r,
model in which only a fraction of the total number of acid/base groups of the protein is affected by the surface and, hence, may contribute to the co-adsorption. An alternative plot is that of the total co-adsorption Ar, + Ar,, vs. uOb (Fig. 61, calculated from Fig. 4 through graphical interpolation, using eqs. (4). We find that, within experimental accuracy, the total ion co-adsorption is a linearly decreasing function of the surface charge density of the blank (the data from linear regression analysis are collected in Table 1). This is a surprising and unexpectedly simple result for such a complicated system. As far as we know, this is the first time the effect has been noted. We will come back to it in more detail in the Discussion below.
Slope
(m’/mC)
(nm’/molecule)
-0.33 -0.27 -0.25
53 44 39
0.97 0.99 0.99
-2.4 - 1.6 -1.9
0.96 0.99 0.98
4.9 1.7 1.8
- 0.26 -0.26 -0.23
42 41 37
0.99 0.98 0.93
5.1 1.3 1.3
-0.24 -0.20 - 0.20
38 33 32
.lG. E. M. Fraaije et al. / Thermodynamics of ion co-adsoption, Table 2 Experimental verification of linkage relation eq. (Sa) a. Conditions: r, = 1.0 mg/m*, pH 5
6 7 8 9 10 average
9.0 8.2 7.0 6.5 5.3 7.2
9.4 8.9 6.0 5.4 4.2 6.8
a Experimental accuracy of the listed differential coefficients is 0.5-l unit; the accuracy of the averages is 0.3-0.4 unit.
and the ion activities a”+ and a%+. Two linkage relations are of special interest because they can experimentally be verified. The relations are, assuming the salt activity is constant:
Equation (5a) was tested by combination of the co-adsorption curves in Figs. 5 for r, = 1.0 mg (pH 4 and 5) and r, = 0.93 (pH 6) by a finite difference scheme, neglecting the small difference in the protein adsorbed amounts (the effect of r, on the co-adsorption numbers is small anyway). The left-hand side (1.h.s.)was calculated for various values of pAg (A@) from the isopH Arab curve for pH 5. The right-hand side (r.h.s.1 was approximated by central difference between the pH 4 and pH 6 curves. The results, shown in Table 2, indicate that the linkage relation eq. (5a) is indeed obeyed; the average values of the 1.h.s and r.h,s do not differ more than 6%, which is about the experimental accuracy. The 1.h.s of eq. (5b) was calculated directry from the ratio of the slopes of the Ar, and Arab curves from Figs. 5 (pH constant). The r.h.s was calculated from combination of detailed data from the pH-stat. Specifically, the 6pH shifts induced by addition of the KI titrant were always small, usually less than 0.1 pH unit. These shifts were
III
273
corrected by the pH-stat, which resulted in turn in small shifts SpAg, usually less than 0.05 pAg unit. The shifts SpAg induced by the pH-stat were always concomitant with very small shifts in Ar, (less then 0.1 co-adsorption unit), such that Ar, z constant. We therefore approximated the differentials apH and apAg in the r.h.s of eq. (5b) by apH and apAg respectively. Closer inspection showed that these detailed data were too inaccurate for test of the linkage relation in each individual point in the titration curves. We performed a statistical analysis of the ratio of the 1.h.s by the r.h.s of eq. (5b) and found that the ratio was approximately normal distributed, indicative of random errors, with an average value of 0.96 and standard deviation of about 1.06 (calculated from 241 pH adjustment cycles). We take this as evidence that eq. (5b) is also obeyed. 4.2 Model interpretation According to the contact layer model 131,the charge regulation is confined to the small layer of ‘atomic’ contact between protein and surface. An additional premise is that the contact layer as a whole is electroneutral. The latter requirement is fulfilled when the protein body electrostatically shields the contact layer from the solution completely. This can be checked by measuring the c-potential in an electrophoresis experiment: if the contact layer is shielded then the surface, fully covered with protein, should titrate as the protein in solution. Some years ago, Norde has shown this to be the case for albumin adsorption on polystyrene [9]. We assume that the same type of shielding mechanism applies also to Albumin adsorption on AgI. In the present case an additional complication has to be considered, namely that the solution side of the adsorbed protein binds Ag+ and/or I-. We measured the Ag+/I- binding capacity of BSA in solution by a few additional titration experiments in the range pH 6-8 and pAg 4-6. Figure 7 shows that the Agf binding rapidly increases with increasing pH and decreasing pAg, indicative of pH-controlled complexation, possibly with amine containing residues [19]. However, under the conditions prevailing in the AgI experi-
J.G.EM Fraaijeet al. / Thermodynamicsof ion co-adsorption,III
274
./
ra 4
I
8
PH I
, 7
Combination of eqs. (1,3,6-S) gives the basic expression for the charge balance in the contact layer: Ar, t Arab = -frfb - a,ut/F
- r,”
(9)
Experimentally it was found (Fig. 6) that Ar, + Arab is a linearly decreasing function of the blank PM Fig. 7. Number P of Ag+ ions bound by BSA in solution, 0.1 M EDJO,, vs. pAg. pH indicated. BSA concentration is 1 mg/ml.
ments the Agf binding may be neglected. The Ibinding was below the detection limit. The implementation of the contact layer model is as follows. The number of surface charges in the contact layer r,’ is the number already present in the blank plus the number of co-adsorbed surface charges: r,C=Ar,ta
4
pF
where aP is the area of the contact layer. Note that the binding and co-adsorption numbers are dimensionless, ap is in m’/mol. We assume that a fraction f of the acid/base groups of the protein is in the contact layer. This fraction is related to the area ap by some geometrical factor, but here we do not need to impose such a relation: f and aP can in principle be estimated separately. The number of excess protons in the contact layer r& is the fraction f of the proton binding number in solution plus the number of co-adsorbed protons: rcab = Arab ffr;
(7)
where rat is the acid/base (proton) binding number in solution. Co-adsorbed electrolyte introduces an additional number of charges r,” in the contact layer. The number rScis defined as the excess number of incorporated K+ over that of NO;:
(8)
surface charge density; this would indicate that f, aP and r,” do not significantly depend on the Galvani potential so that they can be calculated directly from the data from Table 1. A few features are noteworthy: (i) Slope. The values for aP are realistic. The (native) shape of Albumin is that of a proIate spheroid, with dimensions I.7 X 2.7 X 11.6 nm (unhydrated) 1171,or according to another estimate 4 x 4 x 14 nm (hydrated) [18]. The projection area of such an ellipsoid, adsorbed sideways on a surface, would be between 26 nm2 (unhydrated) and 44 nm* (hydrated). This is in the range of the up values listed in Table 1. The higher values of aP, especially compared with the unhydrated native dimensions, indicate to some unfolding. An additional feature is that ap is (slightly) dependent on r,, such that the higher the adsorbed amount, the smaller ap is. A tentative explanation would be that the protein is more unfolded when the surface concentration is low, and that with increasing adsorption the unfolding decreases in order to accommodate more molecules on the surface. There also is the trend that ap is highest for pH 4. Tanford has shown that BSA undergoes reversible expansion below pH 4.3 [16]; such expansion could be facilitated by surface-protein interactions. (ii) Intercept. The accuracy of the intercepts (Table 1) is Iimited (a few co-adsorption units) but it is clear that the intercept is slightly negative for pH 4 and slightly positive for pH 5 and 6. The values for the proton binding numbers in solution are (see above) 29-34 (pH 41, 6-7 (pH 5) and -3 (pH 6). We do not have independent information on f and rScbut a rough estimate for rsc can be obtained as follows. Suppose, for sake of argument, that the shape of the adsorbed protein is that of a flattened ellipsoid, such that the entire part of the molecule which is oriented to the surface is in the contact zone. In this
J.G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption,III
picture f would be about 0.5. Combination of the average values for the intercepts with the proton binding numbers gives: rt 2 - 13 to - 15 (pH 4), - 6 (pH 5), - 4 (pH 6). The fact that these data are negative means that there is a preference for accumulation of anions NO, in the contact layer. It is well known that KNO, is an almost perfect indifferent electrolyte for AgI; the specific interaction of Kf and/or NO; with the AgI surface is very small [lo]. On the other hand, BSA does bind anions specifically [4,16], especially in acid media. It is therefore possible that the protein retains some of its bound NO; when it adsorbs. In this picture, the NO; ions would partly be incorporated in the contact layer, in proportion to the fraction of acid/base groups in the contact layer. Previously [4], we calculated the anion binding capacity of albumin through a thermodynamic interpretation of the Tanford titration data. For Cl- binding in 0.07 M KCI we found: anion binding z 25 (pH 41, 7.5 (pH 5) and - 1 (pH 6). When a fraction f z 0.5of these anions incorporates in the contact layer then r,”= - 12.5 (pH 41, - 3.3 (pH 5) and +OS (pH 6); this is in fair agreement with the values calculated above.
5. Conclusions We have demonstrated that charge regulation in protein adsorption can be quantified experimentally by pH-stat controlled AgI titrations. The experiments show that the proteins exchange reversibly between the adsorbed and dissolved state and that the charge regulation is consistent with thermodynamics. The experimental linear relation between the total ion co-adsorption and the surface charge density of the blank can be interpreted with a simple model, in which the charge regulation is confined to the layer of contact between protein and surface. We did not address model interpretation of the shifts of Galvani potential and interfacial capacitance in relation to the ion co-adsorption. This requires extension of the model to specify exactly where the ions are incorporated in the contact layer. In a following publication we will come back to this problem.
275
Acknowledgements
The research was made possible by financial support from the Netherlands Organization for Scientific Research (NWO). M. Bakkenes and A. Korteweg are acknowledged for technical assistance, G. Buurman and A. Hoekstra for preparing some of the figures.
Appendix
The linkage relations can be derived from a set of coupled Gibbs-Duhem relations for a system in which the surface-protein solution is in membrane equilibrium with its dialysate [2]. Here, we identify the blank as the thermodynamic reference. The differential of the surface pressure 7~’ is (eq. (A71 from [2], for the case c, S- ci):
+T,Ar,
dkto + -
dn
(AI)
CP
where rrs = y - y’, y is the surface tension of the AgI surface (partly) covered with protein, y“ is the surface tension of the blank surface and 7~is the osmotic pressure between the surface-protein solution and its dialysate. In the present case the salt activity is constant (cS= 0.1 M), so that dp.KNo,= 0, d~m,,O, = - dhxon = RT d ln(a,+) and dpLAgNOX = -dpKI -- RT d ln(a,,+). Using the definitions (1, 3, 4) we find: da” = RTT,Ar,, d ln( a”+) + RTI’,Ar,
d ln( aAg+)
r tp
dn.
(A21
cP
The linkage relations eqs. (5a-b) are obtained by a few additional Legendre transformations and subsequent cross-differentiations. Equation (5a) is found from the differential of #- (T,/c,)a and eq. (5b) from the differential of 7’ (T,/c,)a - RTT,Ar, I&,+).
276
J. G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption, Ill
References 1 J.G.E.M. Fraaije, E.-J. Rijnierse, R. Hilhorst and J. Lyklema, Colloid Polym. Sci. 268 (1990) 8.55. 2 J.G.E.M. Fraaije, R.M. Murris, W. Norde and J. Lyklema, Biophys. Chem. 40 (1991) 303. 3 J.G.E.M. Fraaije, W. Norde and J. Lyklema, Biophys. Chem. 40 (1991) 317. 4 J.G.E.M. Fraaije and J. Lyklema, Biophys. Chem. 39 (1991) 31. 5 W. Norde and J. Lyklema, J. Colloid Interface Sci. 66 (1978) 277. 6 A.V. Elgersma, R.L.J. Zsom, W. Norde and J. Lyklema, J. Colloid Interface Sci., 138 (1990) 145. 7 P. van Dulm, W. Norde and J. Lyklema, _I. Colloid Interface Sci. 82 (1981: 77. 8 H. Shirahama, T. Shikuma and T. Suzawa, Colloid Polym. Sci. 267 (1989) 587. 9 W. Norde, Adv. Colloid Interface Sci. 25 (1986) 267. 10 B.H. Bijsterbosch and J. Lyklema, Adv. Colloid Interface Sci. 9 (1978) 147.
11 B.H. Bijsterbosch and J. Lyklema, J. Colloid Interface Sci. 20 (1965) 665. 12 L.K. Koopal and J. Lyklema, Faraday Discuss. Chem. Sot. 59 (1975) 230. 13 A. de Keizer and J. Lyklema, J. Colloid Interface Sci. 75 (1980) 171. 14 H.A. van de1 Schee and J. Lyklema, in: The effect of polymers on dispersion properties, ed. Th.F. Tadros (Academic Press, New York, NY, 1982) p. 81. 15 J. Janatova, J.K. Fuller and M.J. Hunter, J. Biol. Chem. 243 (1968) 3612. 16 ‘C. Tanford, S.A. Swanson and W.S. Shore, J. Am Chem. Sot. 77 (1955) 6421. 17 CL. Ridiford and B.R. Jennings, Biochim. Biophys. Acta 126 (1966) 71. I8 P.G. Squire, P. Moser and CT. O’Konski, Biochem. J., 7 (1968) 426. 19 J. Steinhardt and S. Beykock, in: The proteins, II, ed. J. Neurath (Academic Press, New York, NY, 1964) chap. 8.
263
BIOCHE 01596
Interfacial thermodynamics of protein adsorption and ion co-adsorption. III. Electrochemistry of bovine serum albumin adsorption on silver iodide J.G.E.M. Fraaije *, W. Norde and J. Lyklema Department of Physical and Colloid Chemistry, Agricultural University, Dreijenplein 6, 6703 HB Wageningen (Netherlands)
(Received 7 January 1991;accepted in revised form 19 April 1991)
Abstract An experimental analysis of charge regulation in protein adsorption is presented. The model system consists of colloidal particles of the slightly water soluble salt silver iodide as the adsorbent and the protein bovine serum albumin as the adsorbate. Protein adsorption experiments corroborate earlier findings that albumin adsorbs maximally close to the isoelectric point of the protein. The adsorption is reversible with respect to proteinprotein exchange. The charge regulation is studied by novel potentiometric titrations. The Galvani potential of the adsorbent, partially covered with protein, is varied by the addition of AgNO,/Kl while the pH is kept constant by means of a pH-stat. It is shown that the ion co-adsorption is a linear decreasing function of the blank surface charge density. The results are consistent with thermodynamics: for the first time a few phenomenological linkage relations between the ion co-adsorptions and chemical potentials are verified experimentally. The charge regulation is interpreted in terms of a contact layer model, which explains the ion co-adsorption by compounded ion exchange equilibria in the small layer of atomic contact between adsorbed protein and surface. Keywords: Proteins; Thermodynamics;
Adsorption;
Ion co-adsorption;
1. Introduction
In a series of papers we addressed the interfacial thermodynamics of protein adsorption, ion co-adsorption and related issues. We recently have discussed solubilization of protein in reverse micelles [l], protein ion exchange chromatography [2,3] and the thermodynamics of ion binding
* To whom correspondence should be addressed at AKZO Corporate Research Laboratories Arnhem, Department Applied Mathematics, P.O. Box 9300, 6800 SB Arnhem, Netherlands.
0301-4622/91/$03.50
Ion exchange; Silver iodide; Bovine serum albumin
to proteins in solution [4]. In this report we apply interfacial thermodynamics to a model experimental system, namely adsorption of bovine serum albumin (BSA) on the colloid silver iodide (AgO. The model experimental system is chosen such that the Galvani potential of the surface can be controlled, independently of the pH of the solution. Albumin (human and bovine) adsorption in relation to ion co-adsorption has been studied before, using various techniques: adsorption on porystyrene (ion co-adsorption measurement by electrophoresis [5,6] and radiometry [7]) and hydrophobic copolymers (ion co-adsorption mea-
0 1991 - Elsevier Science Publishers B.V. All rights reserved
264
J.G.E.M. Fraaije et al. / Thermodynamicsof ion co-adsorption,Ill
surement by titration [S]). The experiments show that in many instances ion co-adsorption is not solely determined by electrostatic factors. It is well known that protein adsorption may be accompanied by (partial) unfolding and that the unfolding may lead to a shift in ion binding properties (see review [9J and references therein). Especially albumin is an interesting protein in this respect because it is a labile molecule, which has been shown to unfold on various negative surfaces in roughly the same way [9]. For pH close to the isoelectric point (i.e.p., about pH 4.7), where the protein is Ieast labile in solution, the molecure unfolds minimally, leading to relatively high adsorbed amounts of 2-3 mg/m2. For pH > i.e.p. or pH < i.e.p. the protein is more labile and unfolds to a larger extent, and the more so, the further the pH is away from the i.e.p., leading to relatively smaller adsorbed amounts. The coIloid AgI is one of the classical model systems from colloid chemistry (for general properties see review [lo]>. The colloid can be prepared as a sol or a suspension of precipitates (particle size O.l-10pm). The precipitate lends itself to titration, by adding AgNO, and/or IU solution, without unwanted irreversible additional flocculation. Such titrations are usually analyzed in terms of the surface charge density rO, defined as (eq. [15] from [lo]):
where F is the Faraday constant and ri is the adsorbed amount of i in mol per square meter, a, is in Coulomb per square meter. Note that in the present paper, as in the previous papers [l-4] and [lo], we prefer discussion of interfacial quantities in terms of adsorption of electroneutral components. However, for purpose of interpretation one may identify a, with the excess adsorption of Ag+ ion minus that of I- ion. It is not possible to discriminate between a positive adsorption of one of the components and a negative adsorption of the other. The second experimentally observable, by the use of a AgI electrode, is the silver ion activity, expressed as pAg = -log(a,,+), which can be
related to the interfacial Galvani potential A@ through Nernst’s Law: A@= @a+ T
ln( uAg+)
(4
where a,, is a constant. By convention, the value of G0 is chosen such that A@ is zero in the zero point of charge pAg, of the blank in indifferent electrolyte (KNO,). Silver iodide has been titrated in the presences of various adsorbates, e.g. alcohols [ll], polyvinylalcohols {12],tetralkylammonium nitrate salts [13], oligo- and polypeptides [14]. It is generally found that the adsorbates shift the point of zero charge to lower pAg levels, and that they decrease the differential electric capacitance. The former feature is explained by displacement of oriented surface bound water molecules by hydrophobic moieties of the adsorbates, the latter by reduction of the ratio Es/d, between the dielectric constant E, and thickness d, of the inner double layer part. Here, we are interested in the titration of AgI in the presence of BSA, which may cause an additional effect. Namely, BSA contains weak acid and base side groups so that adsorbed BSA may respond to an imposed shift in Galvani potential by a shift in the acid/base equilibria, i.e. the ion co-adsorptions of acid/base and AgNO,/IU may be dependent on the charge and/or potential of the surface. In our model experiment we monitor the extra uptake or release of acid/base with a pH-stat. The pH-stat introduces two additional observable quantities, namely the pH and the acid/base surface charge density a,,, defined as the excess adsorption of acid over that of base: (3)
As before, the adsorbed amounts are those of the electroneutral components. Negative adsorption of acid is experimentally undistinguishably from positive adsorption of base and vice versa. Bovine serum albumine does not contain hydroxyl binding sites, nor does AgI adsorb acid or base (acid or base binding is negligible for pH G 7, see below). We therefore interpret gab in terms of extra
J.G. E.M. Fraaije et al. / Thermdynamics
of ion co-adsorption, III
265
proton charge (extra with respect to the protein in solution) of the acid/base groups of the adsorbed protein. The AgI titration with pHstat allows us to measure in a single experiment the surface charge of the AgI colloid v0 and the acid/base surface charge density a,,, both as a function of the pAg (or A@) and pH. We analyze the results in two complementary ways. First, the findings are discussed with the thermodynamic method [2]. We will show that various phenomenological linkage relations between the ion co-adsorptions can actually be verified experimentally. Second, the observed behavior is explained with the contact layer model we introduced before [3]. The contact layer model explains ion co-adsorption by compounded ion exchange equilibria in the small layer of ‘atomic’ contact between adsorbed protein and surface. The model allows interpretation of various interesting relations, e.g. that between gO,a,, and the area of the contact layer.
mates of the specific surface has been found before [lo-141; no satisfactory explanation has been given so far. In this report we prefer using the area from the dye adsorption measurement. In the Results section (Section 3) we will discuss this choice in some more detail.
2. Experimental
Typically, O-5 ml of protein solution (containing O-3 mg of BSA in 0.1 M KNO,) were added to 5 ml of we11dispersed AgI precipitate (containing 1.7 g Agl in 0.1 M KNO,) in 10 ml Sybron/ Nalge polycarbonate centrifuge tubes. After the total volumes of the tubes had been adjusted to 10 ml with 0.1 M KNO, solution, the tubes were stoppered with polyethylene caps and vigorously shaken for 10 seconds on a whirl mixer. The tubes were then rotated end over end for 16 hours to reach steady state (more than 90% of the final protein adsorption was already reached in less than a quarter of an hour). Next, the tubes were centrifuged for 20 minutes at 20,000 rpm in a Beckman JA-21 centrifuge equipped with a JA-21 rotor. The protein concentration in the supernatant was determined by measuring the absorbance at 280 nm.
2.1 Preparation of siluer iodide precipitate We closely followed previously published procedures [lo-141. Two batches of AgI precipitate (A and B) were prepared as follows. Two liters of 0.105 M AgNO, were slowly added to 2 1 of 0.1 M KI at a rate of about one drop per 10 seconds. The reaction vessel was shielded from light and continuously flushed with carbondioxide-free nitrogen in order to prevent the formation of silver oxides. After extensive washing with de-ionized water the precipitates were stored at pAg z 6 in the dark under nitrogen atmosphere. Measurement of plateau adsorption of methylene blue yielded for the specific surface areas 0.25 m*/g for precipitate A and 0.30 m2/g or precipitate B. The latter precipitate was used for the titrations and protein plateau adsorption measurements. The point of zero charge of precipitate B was 5.80, very close to the literature value 5.67. From capacitance measurement in the point of zero charge in dilute KNO, (1.1 mM) we calculated a specific surface area of 0.88 m*/g (for method see [lo]). The discrepancy between the two esti-
2.2 Preparation of BSA Bovine serum albumin was purchased from Sigma (type VI, y-globuline and fatty acid free). The number of sulfhydryl groups per protein molecule was initially between 0.3 and 0.4. We increased the sulfhydryl content by a purification method [15]; after purification the number was between 0.85 and 0.88. The sulfhydryl group was subsequently blocked by carboxyamidation, either with normal or radioactive iodoacetamide (Cl41 [ 151.The protein, freed from salt and freeze-dried, was stored at -20°C. 2.3 Adsorption isotherms
2.4 Protein-protein exchange First, 5 ml of unlabelled BSA (2 mg/ml) was added to 25 ml of dispersed AgI precipitate (sorid contents 0.37 gram/ml) in a 50 ml Schott/Duran flask. The mixture was then vigorously stirred for 16 hours using a magnetic stirrer. After the pro-
266
J.G.E.M. Fraaije et al. / Thermodynamics
tein adsorption had been determined, a small amount of radioactive BSA was added, less than 5% of the total amount of protein already present, and the stirring was continued. At selected time intervals 0.5 ml aIiquots were taken from the dispersion and centrifuged in a top desk centrifuge in 1 ml polyethylene vials. The concentration of radioactive BSA was determined with a scintillation counter. The adsorptions of radioactive and unlabelled BSA were calculated from the respective depletions.
of ionco-adsorption,
III
and KI solution in 0.1 M KNO, solution. The effect of pH on the E, values was negligible ( < 1 mV>within the pH range we studied (4 < pH G 71, conversely, the pH drift during the Agl electrode calibration was also negligible (< 0.01 pH unit in unbuffered pH 5.5 solution). The Agl solubility product pS was 16.03-16.06, slightly lower than the literature value 16.07 [lo]. The pH electrode was calibrated with commercial buffers (Titrisol) of pH 4 and pH 7. The pH and AgI electrode calibrations were repeated directly before and after every AgI titration.
2.5 Titration, instrumentation 2.7 Titration, procedure The titrations were carried out in a thermostatted (20 ’ C) Schott titration vessel (volume 1.50ml). The vessel was equipped with four AgI electrodes fitted in one holder, a Schott pH glass electrode and a so-called Van Laar salt bridge (see reference 120 in [lo] for original details) with resistance 700 k0. The reference chamber contained saturated KC1 and a Schott Ag/AgCl reference electrode. The electrode potentials were measured with a multichannel voltage meter (HP 3497A Data Acquisition/Control Unit) after they had been converted to low impedance signals with a home-built impedance transformer. The titrants (AgNO, and KI both 0.01 Iw in 0.1 M KNO,, HNO, and KOH both 0.025 M in 0.1 M KNO,) were added with a Metrohm 655 Dosimat. Homogenization of the concentrated dispersion (about 30% by weight) was achieved by stirring vigorously at the bottom and top of the vessel. The experiments were automated by means of a HP 85 microcomputer. Every five minutes, the computer program collected the data from the electrodes, averaged the signals from the four independent AgI electrodes, decided whether equilibrium had set in (tolerated drifts 0.001 pH/min and 0.001 pAg/min, tolerated spread between the AgI electrodes 0.02 pAg), ordered the burets to add small aliquots of titrant (O.Ol0.1 ml) if necessary, and stored the data. 2.6 Titration, calibration The AgI electrodes were calibrated by short titrations (from 10e5 to 5 . 10e3 M) with AgNO,
First, wet AgI precipitate was added to the titration vessel to a final amount of 28-30 gram, dispersed in 100 ml 0.1 M KNO, and adjusted to pAg 5.5 and selected pH. Then a blank titration was carried out, up to pAg 10.5, in nine steps by successive addition of KI titrant. Next, the titration was reversed in one step by the addition of AgNO, titrant. The initial pAg was always recovered within 0.02 pAg unit. Then 2 mg BSA was added as 1 ml freshly prepared protein solution in 0.1 M KNO, and also at the selected pH level (* 0.001 pH unit). Any pH changes in the titration vessel were first corrected by the pHstat and then the KI titration was repeated. During the titration, the pH-stat continuously monitored the pH, and corrected drifts of pH by addition of the acid and/or the base titrant. In the experiments described, the pH-stat tolerance was set to (in units of pH): 0.001 (pH 4), 0.003 (pH 5) and 0.01 (PH 6). The total procedure for one titration curve took 6-8 hours, the scheme was repeated several times with increasing loads of BSA.
3. Results 3.1 Adsorption isotherms Some typical adsorption isotherms are presented in Fig. 1 as a function of initial pH and pAg. The specific surface area we used for the calculation of the adsorption values in mg/m’
J.G.E.M. Fraaije et al. / Thermodynatnicsof ion co-adsorption,III
was from the dye plateau adsorption measurement (0.25 m*/g for precipitate A), as explained in the experimental section. The protein plateau adsorption values (see also Fig. 2) range from l-2.5 mg/m*, in fair agreement with commonly accepted plateau adsorption values of albumin [9]. If we would have used the surface area (0.88 m*/g) from the capacitance measurement, then the plateau values would have been in the range 0.3-0.8 mg/m’, which is obviously too low. The adsorption isotherms of Fig. 1 reveal that up to about 75% of the plateau adsorption the bulk concentration is below 0.001 mg/ml (detection limit). The isotherms are for four different combinations of initial pH and pAg, including adsorption of (almost) uncharged protein on (almost) uncharged AgI (pH 5, pAg 6.1) and at the other extreme adsorption of negative protein on (very>negative AgI (pH 7, pAg 10.1). The adsorption clearly remains of the high affinity character, even for the case where negative BSA adsorbs on negative AgI. We found that after protein adsorption the equilibrium pH and pAg levels where slightly shifted (less than 1 pH or pAg unit) from the initial values, indicating that the protein adsorption was accompanied by ion co-adsorption. In
267
r, Img /m?
20
!
15-I
10-l
05
L-m-7 PH
Fig. 2. Plateau adsorption of BSA on Agl vs. equilibrium pH and pAg (interpolated, see text) in 0.1 M IWO,.
general, these shifts were similar to those measured more directly in the titration experiments described below. 3.2 F’fateau adsorption
0
I
0.05
,
0.1
I
0.15
I
0.2
I
0.25
cp(mglml) Fig. 1. Adsorption & of BSA on AgI vs. equilibrium concentration cp in 0.1 M IWO,. Initial pAg, pH values: (v ) pAg 6.1, pH 7; (0) pAg 10.1, pH 7; (A) pAg 6.1, pH 5; and (0)
pAg 10.1, pH 5.
In a separate experiment we varied the initial pH and pAg (in a matrix of 11 points variation in pH vs. 4 points variation in pAg, i.e. in total 44 points), while the total amount of protein was held constant (2.5 mg>. The equilibrium protein concentration in bulk varied with adsorbed amount from 0.11 mg/ml to 0.2 mg/ml, this is in the plateau regime (Fig. 1). These adsorption values were (graphically) interpolated with respect to the equilibrium pH and pAg values, to values for constant pH and pAg. The results are shown in Fig. 2. The curves show that the adsorption is a maximum at pH 4, close to, but siighly lower than the i.e.p. We noted already in the Introduction that Albumin adsorbs on a variety of negative surfaces with maxima close to the i.e.p. of the protein (pH 4.7) [91, our results confirm this. There is a small effect of the surface charge:
268
J.G.E.M. Fraaije et al. / Thermodynamics
of ion co-adsorption, IfI
_________________--------k
,OO
*L
per, %
O
@@A-A-
6060-
/
'o?&+&?d&@~ ipO&----o zo-
0
I
I I20
60
1 240
1SO
I 420
I 360
I 300
n y
I 1420
t(mn)
Fig. 3. Protein exchange ratio p.e.r. vs. exchange time. 0.01 M KNO,, pAg 10: (0) pH 6, cp = 0.08 mg/ml; mg/ml; ( a) pH 4, cP = 0.03 mg/ml.
we find that the adsorption is slightly higher when the charges of protein and surface are opposite.
rapid exchange to p.e.r. E 30-40% in the first few minutes, and a slower exchange to p.e.r. E 100% in the following hours. Moreover, there is Little influence of the pH. This demonstration of reversibility is important: it is a prerequisite to verify the thermodynamic linkage relations in the Discussion.
3.2 Protein-protein exchange experiments As a measure of the protein exchange we used the protein exchange ratio (p.e.r.) = (r/C)radioactive BSA ' (C/r)u"labelled BSA' The
(x) pH 5, cp = 0.07
3.3 Titrations
de-
The titration curves are shown in Fig. 4. Adsorption of the protein is reflected in the shifts in u,, and a,, with respect to the blank values. Note
pendence of p.e.r. on the medium conditions and time is depicted in Fig. 3. The results clearly indicate that the adsorption is reversible, with a
-100(a)
1 I
pH4
I
I
I
I
/
1
I
I
I
I
I
I
6
7
6
9
10
11
6
7
6
9
10
11
PAg
PAg
Fig. 4. Titrationof A&BSA complex in 0.1 M KNO, at fixed pH, r0 is surface charge density of AgI (eq. [l]), vaah is acid/base surface charge density (eq. [3]). pH 4 and pH 5: (0) blank; (x) r,,= 0.47 mg/m2; (a) rP = 0.93 mg/m2; and (0) rP = 1.4 mg/m’,
ph 6: (0) blank; (X) r, = 0.33 mg/m’;
(A) r, = 0.67 mg/m*; and (u) r, = 0.93 mg/m*.
J.G.E. M. Fraaije et al. / Thermodynamicsof ion co-adsorption, Ill
u,JmC/m*
269
1
soo - (4
I 6
7
e
9
PA9
I IO
I 6
I 11
I 7
a
9
PA9
IO
J 11
Fig. 4. (continued).
that in these experiments the protein adsorption levels are beIow the plateau adsorption values and that consequently the protein concentration in solution is negligibly low. We may therefore identify the charge regulation with the response of adsorbed protein only and neglect contributions from titration of the protein in solution. The blank curves of a0 for the different pH levels are identical, illustrating the absence of base/acid adsorption on the bare AgI surface. This is confirmed by the (undetectably) low
acid/base adsorption craabin the blank titration, The shapes of the blank r0 curves are in good agreement with published titration curves of Agl precipitates in 0.1 M KNO,, although the absolute values from our measurements are about a factor 3 higher [lo]. The explanation lies in the discrepancy between the specific surface area from capacitance measurement (often used in the older work [lo], 0.88 m2/g for the precipitate B), whereas here we prefer to use the specific surface area from the dye adsorption measurement (0.30
ff.b(mC/m*
)
$100 if) t
PH 6
Fig. 4. (continued).
J.G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption, Iii
270
*r,
40
(a)
J
.1 -0.3
*
4c
40
-0.2
-0.1
0
*WV)
r,,
(4
(C)
x pH5
B
2c
A
f
t-
C
pH5
-at
-0.3
-03
I
-20
0
13
/
-a2
A+(V)
,I
-0.1 *O(V)
Ar.
*r,,
40
4C
(4
2c
(
20
t
01
0
B
*
a
P%
-20
1
-0.3
-92
-01 AOCV)
-“.a
-a2
-01
0
Ablv)
Fig. 5. Ion co-adsorption numbers Ari vs. interfacial Galvani potential A@. pH 4 and pH 5: (0) blank; (*) r = 0.47 mg/m2; (A) = 0.93 mg/m*); and (0) r, = 1.4 mg/m’. pH 6: (0) blank; (*I r, = 0.33 mg/m’; (A) r, = 0.67 mg/mP; and ( q ) T, = 0.93 mg/m*.
r,
J.G.E.M. Fruaije et ul. / Thermodynamicsof ion co-adsorption,III
m2/g for the precipitate B). Note that we do so in order to make the BSA plateau adsorption levels conform to the literature values. We find that in general the more negatively charged protein (pH 6) makes a0 more positive, and that the more positively charged protein (pH 4) makes a, more negative (except for pAg > 8.5). Also, in all cases a more negatively charged surface makes the protein more positive. These effects are clearly a result of electrostatic induction. Note however that for pH 4 (Fig. 4a) there is a common intersection point (c.i.p.) in the a, curves (around pAg 8.51, such that for pAg > c.i.p. the surface, partially covered with positively charged protein, is less negatively charged than the blank. A second feature is that the differential electric capacitance C (= b&IA@ = - 17.19 auJ apAg in F/m’> is not much affected by the protein adsorption. Comparison with effects of other types of adsorbates [lo-141 shows that in this respect BSA behaves exceptionally (see Introduction). In general, the effect of an adsorbate is to reduce C significantly, which is explained by a possible increase in the thickness of the Stern layer and/or lowering of relative dielectric constant of the Stern layer. The implicit assumption is then that the Stern layer itself cannot adjust in charge, i.e. is not titratable. Here, the adsorbed protein readily adjusts its charge through titration of the acid/base groups in the contact zone between protein and surface, which is the region dominating C.
271
(a) pH4
-120
-80
-40
40
0
dcmom?,
b) PHI
-120
-80
-40
0
dhlC/m2) A& + Ar,,
4or
3.4 ion co-adsorption numbers A slightly different presentation of the titration results, especially suited for the thermodynamic and model interpretation, is in terms of the ion co-adsorption numbers of the acid/base, Arab, and AgNO,/KI, Ar,. These are the numbers of ions extra adsorbed or desorbed per adsorbed protein molecule. They are defined through: 1 Ar,=I;T(q,-qb) P
Fig. 6. Total ion co-adsorption Ar, + Ar,, vs. surface charge density of the blank I$. Symbols as in Fig. 5.
1. G.E.M. Fraaije et al. / Thermodynamicsof ion co-adsorption,III
272
where gob is the surface charge density of the blank. The co-adsorption numbers were calculated from the original titration curves Figs. 4 through graphical interpolation. The results are shown in Figs. 5 (note that we have also converted pAg to the Galvani potential by the use of eq. [2] and pAg, = 5.67 for the blank). We recall from the proton titration experiments by Tanford [121 that the isoionic point of BSA in solution is about pH 5.5-5.6 and the isoelectric point is about 4.7. The proton binding number is about 29-34 for pH 4, 6-7 for pH 5 and -3 for pH 6 (lower numbers measured in 0.08 M KCI, higher numbers in 0.15 M KCl>. The maximum positive proton binding number is about 96, i.e. this number of protons may be bound in excess to the isoionic state [4]. This implies that the maximum number the protein can co-adsorb is about 62-67 at pH 4, 89-90 at pH 5 and 99 at pH 6. The co-adsorption curves show that in the present case the co-adsorption levels are far below the maximally attainable levels, indicating that in principle the protein has ample capacity to adjust in charge by the extra uptake or release of protons. However, it is possible that the saturation observed near Arab G 30 for pH 4-5 (Figs. 5b,d) indicates that the effective maximally attainable co-adsorption numbers are smaller than the theoretical values we calculated. This may be explained by the contact layer
Table
PHI 0.47 0.93 1.4 PHI 0.33 0.67 1.0
4. I Thermodynamic interpretation
Generally speaking, thermodynamics predicts that co-adsorption numbers and chemical potentials are related through various Maxwell (or linkage) equations. These equations arc easily dcrived from a set of coupled Gibbs-Duhem relations, as we have shown in [2]. In the Appendix the theory is applied to find the linkage relations between the co-adsorption numbers Ar, and Arab
analysis of Ar, + Ar,, vs. uCv(Fig. 6) R2
Intercept
(ms/m*) pH4 0.47 0.93 1.4
4. Discussion
1
Regression
r,
model in which only a fraction of the total number of acid/base groups of the protein is affected by the surface and, hence, may contribute to the co-adsorption. An alternative plot is that of the total co-adsorption Ar, + Ar,, vs. uOb (Fig. 61, calculated from Fig. 4 through graphical interpolation, using eqs. (4). We find that, within experimental accuracy, the total ion co-adsorption is a linearly decreasing function of the surface charge density of the blank (the data from linear regression analysis are collected in Table 1). This is a surprising and unexpectedly simple result for such a complicated system. As far as we know, this is the first time the effect has been noted. We will come back to it in more detail in the Discussion below.
Slope
(m’/mC)
(nm’/molecule)
-0.33 -0.27 -0.25
53 44 39
0.97 0.99 0.99
-2.4 - 1.6 -1.9
0.96 0.99 0.98
4.9 1.7 1.8
- 0.26 -0.26 -0.23
42 41 37
0.99 0.98 0.93
5.1 1.3 1.3
-0.24 -0.20 - 0.20
38 33 32
.lG. E. M. Fraaije et al. / Thermodynamics of ion co-adsoption, Table 2 Experimental verification of linkage relation eq. (Sa) a. Conditions: r, = 1.0 mg/m*, pH 5
6 7 8 9 10 average
9.0 8.2 7.0 6.5 5.3 7.2
9.4 8.9 6.0 5.4 4.2 6.8
a Experimental accuracy of the listed differential coefficients is 0.5-l unit; the accuracy of the averages is 0.3-0.4 unit.
and the ion activities a”+ and a%+. Two linkage relations are of special interest because they can experimentally be verified. The relations are, assuming the salt activity is constant:
Equation (5a) was tested by combination of the co-adsorption curves in Figs. 5 for r, = 1.0 mg (pH 4 and 5) and r, = 0.93 (pH 6) by a finite difference scheme, neglecting the small difference in the protein adsorbed amounts (the effect of r, on the co-adsorption numbers is small anyway). The left-hand side (1.h.s.)was calculated for various values of pAg (A@) from the isopH Arab curve for pH 5. The right-hand side (r.h.s.1 was approximated by central difference between the pH 4 and pH 6 curves. The results, shown in Table 2, indicate that the linkage relation eq. (5a) is indeed obeyed; the average values of the 1.h.s and r.h,s do not differ more than 6%, which is about the experimental accuracy. The 1.h.s of eq. (5b) was calculated directry from the ratio of the slopes of the Ar, and Arab curves from Figs. 5 (pH constant). The r.h.s was calculated from combination of detailed data from the pH-stat. Specifically, the 6pH shifts induced by addition of the KI titrant were always small, usually less than 0.1 pH unit. These shifts were
III
273
corrected by the pH-stat, which resulted in turn in small shifts SpAg, usually less than 0.05 pAg unit. The shifts SpAg induced by the pH-stat were always concomitant with very small shifts in Ar, (less then 0.1 co-adsorption unit), such that Ar, z constant. We therefore approximated the differentials apH and apAg in the r.h.s of eq. (5b) by apH and apAg respectively. Closer inspection showed that these detailed data were too inaccurate for test of the linkage relation in each individual point in the titration curves. We performed a statistical analysis of the ratio of the 1.h.s by the r.h.s of eq. (5b) and found that the ratio was approximately normal distributed, indicative of random errors, with an average value of 0.96 and standard deviation of about 1.06 (calculated from 241 pH adjustment cycles). We take this as evidence that eq. (5b) is also obeyed. 4.2 Model interpretation According to the contact layer model 131,the charge regulation is confined to the small layer of ‘atomic’ contact between protein and surface. An additional premise is that the contact layer as a whole is electroneutral. The latter requirement is fulfilled when the protein body electrostatically shields the contact layer from the solution completely. This can be checked by measuring the c-potential in an electrophoresis experiment: if the contact layer is shielded then the surface, fully covered with protein, should titrate as the protein in solution. Some years ago, Norde has shown this to be the case for albumin adsorption on polystyrene [9]. We assume that the same type of shielding mechanism applies also to Albumin adsorption on AgI. In the present case an additional complication has to be considered, namely that the solution side of the adsorbed protein binds Ag+ and/or I-. We measured the Ag+/I- binding capacity of BSA in solution by a few additional titration experiments in the range pH 6-8 and pAg 4-6. Figure 7 shows that the Agf binding rapidly increases with increasing pH and decreasing pAg, indicative of pH-controlled complexation, possibly with amine containing residues [19]. However, under the conditions prevailing in the AgI experi-
J.G.EM Fraaijeet al. / Thermodynamicsof ion co-adsorption,III
274
./
ra 4
I
8
PH I
, 7
Combination of eqs. (1,3,6-S) gives the basic expression for the charge balance in the contact layer: Ar, t Arab = -frfb - a,ut/F
- r,”
(9)
Experimentally it was found (Fig. 6) that Ar, + Arab is a linearly decreasing function of the blank PM Fig. 7. Number P of Ag+ ions bound by BSA in solution, 0.1 M EDJO,, vs. pAg. pH indicated. BSA concentration is 1 mg/ml.
ments the Agf binding may be neglected. The Ibinding was below the detection limit. The implementation of the contact layer model is as follows. The number of surface charges in the contact layer r,’ is the number already present in the blank plus the number of co-adsorbed surface charges: r,C=Ar,ta
4
pF
where aP is the area of the contact layer. Note that the binding and co-adsorption numbers are dimensionless, ap is in m’/mol. We assume that a fraction f of the acid/base groups of the protein is in the contact layer. This fraction is related to the area ap by some geometrical factor, but here we do not need to impose such a relation: f and aP can in principle be estimated separately. The number of excess protons in the contact layer r& is the fraction f of the proton binding number in solution plus the number of co-adsorbed protons: rcab = Arab ffr;
(7)
where rat is the acid/base (proton) binding number in solution. Co-adsorbed electrolyte introduces an additional number of charges r,” in the contact layer. The number rScis defined as the excess number of incorporated K+ over that of NO;:
(8)
surface charge density; this would indicate that f, aP and r,” do not significantly depend on the Galvani potential so that they can be calculated directly from the data from Table 1. A few features are noteworthy: (i) Slope. The values for aP are realistic. The (native) shape of Albumin is that of a proIate spheroid, with dimensions I.7 X 2.7 X 11.6 nm (unhydrated) 1171,or according to another estimate 4 x 4 x 14 nm (hydrated) [18]. The projection area of such an ellipsoid, adsorbed sideways on a surface, would be between 26 nm2 (unhydrated) and 44 nm* (hydrated). This is in the range of the up values listed in Table 1. The higher values of aP, especially compared with the unhydrated native dimensions, indicate to some unfolding. An additional feature is that ap is (slightly) dependent on r,, such that the higher the adsorbed amount, the smaller ap is. A tentative explanation would be that the protein is more unfolded when the surface concentration is low, and that with increasing adsorption the unfolding decreases in order to accommodate more molecules on the surface. There also is the trend that ap is highest for pH 4. Tanford has shown that BSA undergoes reversible expansion below pH 4.3 [16]; such expansion could be facilitated by surface-protein interactions. (ii) Intercept. The accuracy of the intercepts (Table 1) is Iimited (a few co-adsorption units) but it is clear that the intercept is slightly negative for pH 4 and slightly positive for pH 5 and 6. The values for the proton binding numbers in solution are (see above) 29-34 (pH 41, 6-7 (pH 5) and -3 (pH 6). We do not have independent information on f and rScbut a rough estimate for rsc can be obtained as follows. Suppose, for sake of argument, that the shape of the adsorbed protein is that of a flattened ellipsoid, such that the entire part of the molecule which is oriented to the surface is in the contact zone. In this
J.G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption,III
picture f would be about 0.5. Combination of the average values for the intercepts with the proton binding numbers gives: rt 2 - 13 to - 15 (pH 4), - 6 (pH 5), - 4 (pH 6). The fact that these data are negative means that there is a preference for accumulation of anions NO, in the contact layer. It is well known that KNO, is an almost perfect indifferent electrolyte for AgI; the specific interaction of Kf and/or NO; with the AgI surface is very small [lo]. On the other hand, BSA does bind anions specifically [4,16], especially in acid media. It is therefore possible that the protein retains some of its bound NO; when it adsorbs. In this picture, the NO; ions would partly be incorporated in the contact layer, in proportion to the fraction of acid/base groups in the contact layer. Previously [4], we calculated the anion binding capacity of albumin through a thermodynamic interpretation of the Tanford titration data. For Cl- binding in 0.07 M KCI we found: anion binding z 25 (pH 41, 7.5 (pH 5) and - 1 (pH 6). When a fraction f z 0.5of these anions incorporates in the contact layer then r,”= - 12.5 (pH 41, - 3.3 (pH 5) and +OS (pH 6); this is in fair agreement with the values calculated above.
5. Conclusions We have demonstrated that charge regulation in protein adsorption can be quantified experimentally by pH-stat controlled AgI titrations. The experiments show that the proteins exchange reversibly between the adsorbed and dissolved state and that the charge regulation is consistent with thermodynamics. The experimental linear relation between the total ion co-adsorption and the surface charge density of the blank can be interpreted with a simple model, in which the charge regulation is confined to the layer of contact between protein and surface. We did not address model interpretation of the shifts of Galvani potential and interfacial capacitance in relation to the ion co-adsorption. This requires extension of the model to specify exactly where the ions are incorporated in the contact layer. In a following publication we will come back to this problem.
275
Acknowledgements
The research was made possible by financial support from the Netherlands Organization for Scientific Research (NWO). M. Bakkenes and A. Korteweg are acknowledged for technical assistance, G. Buurman and A. Hoekstra for preparing some of the figures.
Appendix
The linkage relations can be derived from a set of coupled Gibbs-Duhem relations for a system in which the surface-protein solution is in membrane equilibrium with its dialysate [2]. Here, we identify the blank as the thermodynamic reference. The differential of the surface pressure 7~’ is (eq. (A71 from [2], for the case c, S- ci):
+T,Ar,
dkto + -
dn
(AI)
CP
where rrs = y - y’, y is the surface tension of the AgI surface (partly) covered with protein, y“ is the surface tension of the blank surface and 7~is the osmotic pressure between the surface-protein solution and its dialysate. In the present case the salt activity is constant (cS= 0.1 M), so that dp.KNo,= 0, d~m,,O, = - dhxon = RT d ln(a,+) and dpLAgNOX = -dpKI -- RT d ln(a,,+). Using the definitions (1, 3, 4) we find: da” = RTT,Ar,, d ln( a”+) + RTI’,Ar,
d ln( aAg+)
r tp
dn.
(A21
cP
The linkage relations eqs. (5a-b) are obtained by a few additional Legendre transformations and subsequent cross-differentiations. Equation (5a) is found from the differential of #- (T,/c,)a and eq. (5b) from the differential of 7’ (T,/c,)a - RTT,Ar, I&,+).
276
J. G.E.M. Fraaije et al. / Thermodynamics of ion co-adsorption, Ill
References 1 J.G.E.M. Fraaije, E.-J. Rijnierse, R. Hilhorst and J. Lyklema, Colloid Polym. Sci. 268 (1990) 8.55. 2 J.G.E.M. Fraaije, R.M. Murris, W. Norde and J. Lyklema, Biophys. Chem. 40 (1991) 303. 3 J.G.E.M. Fraaije, W. Norde and J. Lyklema, Biophys. Chem. 40 (1991) 317. 4 J.G.E.M. Fraaije and J. Lyklema, Biophys. Chem. 39 (1991) 31. 5 W. Norde and J. Lyklema, J. Colloid Interface Sci. 66 (1978) 277. 6 A.V. Elgersma, R.L.J. Zsom, W. Norde and J. Lyklema, J. Colloid Interface Sci., 138 (1990) 145. 7 P. van Dulm, W. Norde and J. Lyklema, _I. Colloid Interface Sci. 82 (1981: 77. 8 H. Shirahama, T. Shikuma and T. Suzawa, Colloid Polym. Sci. 267 (1989) 587. 9 W. Norde, Adv. Colloid Interface Sci. 25 (1986) 267. 10 B.H. Bijsterbosch and J. Lyklema, Adv. Colloid Interface Sci. 9 (1978) 147.
11 B.H. Bijsterbosch and J. Lyklema, J. Colloid Interface Sci. 20 (1965) 665. 12 L.K. Koopal and J. Lyklema, Faraday Discuss. Chem. Sot. 59 (1975) 230. 13 A. de Keizer and J. Lyklema, J. Colloid Interface Sci. 75 (1980) 171. 14 H.A. van de1 Schee and J. Lyklema, in: The effect of polymers on dispersion properties, ed. Th.F. Tadros (Academic Press, New York, NY, 1982) p. 81. 15 J. Janatova, J.K. Fuller and M.J. Hunter, J. Biol. Chem. 243 (1968) 3612. 16 ‘C. Tanford, S.A. Swanson and W.S. Shore, J. Am Chem. Sot. 77 (1955) 6421. 17 CL. Ridiford and B.R. Jennings, Biochim. Biophys. Acta 126 (1966) 71. I8 P.G. Squire, P. Moser and CT. O’Konski, Biochem. J., 7 (1968) 426. 19 J. Steinhardt and S. Beykock, in: The proteins, II, ed. J. Neurath (Academic Press, New York, NY, 1964) chap. 8.
